Presentation | 2015-05-15 Improving Generalization Ability of SpikeProp Networks Takuya Toyota, Yoichiro Kitagawa, Haruhiko Takase, Hiroharu Kawanaka, Shinji Tsuruoka, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Recently, spiking neural networks attracts many researchers attention as an artificial neural network model that has an ability to time series information processing. In this article, we focus on SpikeProp, which is a kind of spiking neural networks. It often degrades generalization ability. We discuss a method to detect inputs that cause the problem. In our previous research, we can detect signs of degradation by observing activity of each unit: (1) there is a local maximum of the activity of an unit around the threshold, (2) the activity of an unit stays around threshold for long time. We show results of the dependency on its parameters. The proposed method judged 85% of teacher patterns correctly: whether it causes the problem, or not. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Spiking neural network / SpikeProp / Generalization |
Paper # | SIP2015-13,IE2015-13,PRMU2015-13,MI2015-13 |
Date of Issue | 2015-05-07 (SIP, IE, PRMU, MI) |
Conference Information | |
Committee | PRMU / MI / IE / SIP |
---|---|
Conference Date | 2015/5/14(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | Kazuhiko Sumi(Aoyama Gakuin Univ.) / Akinobu Shimizu(Tokyo Univ. of Agric. and Tech.) / Toshiaki Fujii(Nagoya Univ.) / Yoshinobu Kajikawa(Kansai Univ.) |
Vice Chair | Koichi Kise(Osaka Pref. Univ.) / Shuji Senda(NEC) / Yoshitaka Masutani(Hiroshima City Univ.) / Kensaku Mori(Nagoya Univ.) / Seishi Takamura(NTT) / Takayuki Hamamoto(Tokyo Univ. of Science) / Osamu Houshuyama(NEC) / Makoto Nakashizuka(Chiba Inst. of Tech.) |
Secretary | Koichi Kise(Kyushu Univ.) / Shuji Senda(Omron) / Yoshitaka Masutani(Tokushima Univ.) / Kensaku Mori(Kinki Univ.) / Seishi Takamura(NHK) / Takayuki Hamamoto(KDDI R&D Labs.) / Osamu Houshuyama(Ritsumeikan Univ.) / Makoto Nakashizuka(NEC) |
Assistant | Wataru Ohyama(Mie Univ.) / Mitsuru Anbai(DENSO IT Lab.) / Takayuki Kitasaka(Aichi Inst. of Tech.) / Hidetaka Hontani(Nagoya Inst. of Tech.) / Shohei Matsuo(NTT) / Takamichi Miyata(Chiba Inst. of Tech.) / Takamichi Miyata(Chiba Inst. of Tech.) |
Paper Information | |
Registration To | Technical Committee on Pattern Recognition and Media Understanding / Technical Committee on Medical Imaging / Technical Committee on Image Engineering / Technical Committee on Signal Processing |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Improving Generalization Ability of SpikeProp Networks |
Sub Title (in English) | Method to Detect Inputs that Cause Problem |
Keyword(1) | Spiking neural network |
Keyword(2) | SpikeProp |
Keyword(3) | Generalization |
1st Author's Name | Takuya Toyota |
1st Author's Affiliation | Mie University(Mie Univ.) |
2nd Author's Name | Yoichiro Kitagawa |
2nd Author's Affiliation | Mie University(Mie Univ.) |
3rd Author's Name | Haruhiko Takase |
3rd Author's Affiliation | Mie University(Mie Univ.) |
4th Author's Name | Hiroharu Kawanaka |
4th Author's Affiliation | Mie University(Mie Univ.) |
5th Author's Name | Shinji Tsuruoka |
5th Author's Affiliation | Mie University(Mie Univ.) |
Date | 2015-05-15 |
Paper # | SIP2015-13,IE2015-13,PRMU2015-13,MI2015-13 |
Volume (vol) | vol.115 |
Number (no) | SIP-22,IE-23,PRMU-24,MI-25 |
Page | pp.pp.65-70(SIP), pp.65-70(IE), pp.65-70(PRMU), pp.65-70(MI), |
#Pages | 6 |
Date of Issue | 2015-05-07 (SIP, IE, PRMU, MI) |